Posts

Back in 2015, I was debugging a SQL Server cluster that worked perfectly. Except when it didn’t. Connections timed out randomly, failovers failed silently, and nothing made sense. I was sure it was DNS. It was DNS. But that wasn’t the real problem. The real problem was me, seeing exactly what I expected to see instead of what was actually there. Turns out, pilots and DBAs have something important in common: expectation bias can really mess you up.

You did the hard work. Found the buried metric, rebuilt the report, put the lever in the top-left tile. The meeting went beautifully. Everyone nodded. Then nothing changed. Here’s what I keep relearning: a lever is not a mechanism. It’s a person with a calendar, a bonus structure, and reasons to leave things exactly as they are. Most reports die in the gap between who requested them and who has to work differently on Monday.

Every student pilot learns this the hard way: staring at the altimeter won’t fix your altitude. It’s a readout, not a control. Your dashboards have the same problem. Most of the metrics sitting at top left are gauges, things people can only watch. The actual levers, the numbers someone could move on Monday morning, are buried three clicks deep rendered in grey font. Three questions tell you which is which. The answer is usually uncomfortable.

Software versioning is solved. Ontology versioning isn’t. When you change what counts as an ‘active customer,’ that’s not a schema migration with a clear notion of correctness. It’s a contested definition where reasonable people can land in different places. Fabric IQ gives you the artifact, but it doesn’t ship the governance workflow. That part you build yourself. This post digs into what running layered ontologies actually takes, and why the organizational problem matters more than the technical one.

A sharp LinkedIn critique argues declared ontologies are broken by design. Model your business formally, and by the time you’re done, it’s already changed underneath you. The proposed fix: let structure emerge from data instead of declaring it upfront. It’s a compelling pitch, but I think it misdiagnoses the problem. Ontology drift isn’t evidence that formalization failed. It’s evidence we failed to separate what should be stable from what shouldn’t. The real answer might be simpler than either camp admits.

I said my rituals are for me, not the audience. That’s technically true. But a ritual that changes your internal state doesn’t keep that change contained. It shows up in your posture, your pace, whether your eyes land on people or drift past them. When your body says one thing while your words say another, the audience isn’t hearing confidence. They’re hearing the gap. That costs you trust and attention, and the science on why got interesting.

You built the pipeline everyone recommends. One model drafts, another reviews. Two sets of eyes. Except your reviewer is a language model, and language models can be argued out of a position by nothing more than a confident counter-argument. No payload to catch. No injection required. Just a wrong document in your retrieval set making a fluent case. The verdict flips, the badge says ‘reviewed,’ and everyone downstream stops looking. You added a laundering step.

My Data Platform MVP renewed for another year, nearly a decade now. I should feel celebratory. Mostly I feel like I owe somebody money. Because none of it was mine to begin with. Every door was opened by someone who didn’t have to. But the compact that built this community, where you give it away and trust it comes back around, I’m not sure it holds anymore. We built the library. Someone else started charging admission.

I spent the winter teaching people to defend against prompt injection. Layer your defenses, I said, and you can ship systems you trust. I still believe that. But I found an attack that walks through every layer I described, and it has no payload at all. It is just an argument. You push back on the model’s conclusion with confidence, and it folds. Every time. No guardrail fires because no guardrail was watching the conclusion itself. That changes things.

We’ve spent years encoding business knowledge into Power BI semantic models. What a customer is, what revenue means. The problem is that knowledge is locked in DAX, invisible to AI agents. Fabric IQ introduces ontologies as the fix, a layer that captures meaning in a form machines can reason against. But generating an ontology from your existing semantic model inherits all its limitations. The real question is whether organisations will do the hard work of agreeing on definitions.

For years I thought dimensional models were about organizing data and making queries fast. That’s true, but it’s profoundly incomplete. Dimensional models describe how we store facts. They don’t describe what those facts mean. That gap shows up the moment someone asks a question your star schema wasn’t designed for. Ontologies solve a different problem: formal, machine-readable definitions of business concepts and their relationships. With Microsoft now shipping Fabric IQ, this conversation isn’t academic anymore.

Describe your chosen architecture to your AI assistant and ask what it thinks. Odds are, it’ll tell you the approach is sound. But a 2025 Stanford study found AI models affirm users 47% more than humans do, even when the user is clearly wrong. Worse: people who got sycophantic responses trusted the AI more and were more likely to return. The version that damaged their judgment was the one they liked best. This is Goodhart’s Law in your feedback loop.

Your AI policy was written for someone who doesn’t work on your team.Your AI policy was written for someone who doesn’t work on your team. Probably for someone who doesn’t exist. Part 2 of this series moves from research to practice. The key variable isn’t cognitive profile — it’s domain expertise asymmetry. Where that gap is largest, the agreement machine runs without a check. This post covers where the real risk concentrates, why structured review consistently outperforms ‘does anyone see any problems?’, and what a policy that actually changes behaviour looks like. Design the workflow. Not the person.

There’s a fundamental difference between cloud and on-prem work that I keep coming back to. Cloud is a sophisticated model kit with pre-designed pieces that snap together. On-prem was a pile of Lego bricks and a problem to solve. Everything in between was yours to figure out. If I had to go back tomorrow, I know exactly what skill would need the most work. It’s not the technical stuff you’d guess. It’s something more fundamental, something that years of managed services have slowly let atrophy.

Your brain evolved to detect lions. Now it may fire every time you open ChatGPT. This post unpacks three reasons LLMs are not neutral tools — what they’re trained on, how RLHF creates systematic pressure toward validation, and what your neurobiology does with the result. Then it gets specific: the same sycophantic system creates meaningfully different failure modes depending on who is using it. Nobody designed this. Nobody fully planned for it. And most deployment practices still aren’t accounting for it. The research is early. The mechanisms are not.

This week I’m heading to Prague for DataPoint, organized by my friend Štěpán Rešl, a man whose brain capacity I genuinely cannot account for. Quick trip: arrive Thursday, deliver my data literacy session Friday morning, fly straight home. No time to explore the city, which I’ve been promised is beautiful. I can confirm the food is extraordinary. I’ve been thinking about one particular meal since 2019. New post on what I’m delivering and why the topic still matters.

Data sovereignty concerns - driven by the CLOUD Act and NIS2 obligations - are pushing Swedish organizations to reconsider on-premises analytics. The problem: a decade of cloud adoption has quietly eroded the skills needed to execute on-prem projects. The Microsoft stack in 2026 is more capable than most assume, but lacks a Fabric equivalent. The real decision is operational ownership, not technology preference. Organizations that can reason across the full cloud-to-on-prem spectrum will navigate the uncertainty best.

The tech industry is cutting junior roles at record pace, calling it AI efficiency. But here’s the problem: you can’t skip a generation of practitioners and expect institutional knowledge to survive. The seniors who remain were once juniors who had space to build understanding from the ground up. That pipeline is being shut off. Part 2 is about what to do about it. How to mentor like you prompt. How to train for reasoning, not just execution - and why this might be the best time to hire juniors, not the worst.

A Wharton study put 1,372 people in front of a deliberately wrong AI. 73% followed it anyway. Not from carelessness - from fluency, confidence, and the absence of a trained skeptical reflex. We built that gap ourselves: ten years of certifications that taught procedure and skipped judgment. Now the procedures belong to agents, and the engineers we trained have no independent model to cross-check against. This is about what we teach instead - and why broken things are the best place to start.

Most speakers spend weeks building a presentation and zero minutes thinking about how to end it. Then they hand the microphone to the audience and call it a close. That’s not a close - it’s an abdication. The serial position effect means your ending is the thing most likely to survive in memory. This post breaks down why Q&A endings undermine you, why stories work at a neurological level, and what the research actually recommends instead.